JP3364827B2 - Audio encoding method, audio decoding method, audio encoding / decoding method, and devices therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To evade deterioration in a regenerative signal and to make possible reproducing a voice with less transmission errors by constituting so that a noise code is not incorporated in the code used when a characteristic position for pitch synchronizing a noise code vector is obtained. SOLUTION: Distances between coded voices generated using adaptive codes, the noise codes, gains respectively selected in the case when an adaptive code book 11 and a noise code book 12 are used and the case when a fixed code book 24 and a second noise code book 25 and an input voice S1 are compared with each other, and the adaptive code, the noise code, the gain that the distance becomes small are selected. After the coding is ended, the codes of the adaptive code, the noise code and the gain minimizing the distortion between the input voice and the coded voice are outputted as a coded result S2. In such a manner, the voice is obtained from the coded code without directly coding the information of the characteristic position, and this method is constituted so that the noise code for the noise code book generating at least a periodic time sequence vector is not incorporated in the code used when the characteristic position is obtained.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice coding / decoding method for compressing and coding a voice signal into a digital signal, and particularly to a voice for reproducing a voice whose quality is less deteriorated by a transmission line error when applied to communication. The present invention relates to a coding method, a voice decoding method, and a voice coding / decoding method.

[0002]

2. Description of the Related Art Conventionally, as a high-efficiency speech coding method for communication, code-excited line coding (Code-Excited Linea) has been used.
Methods such as r Prediction coding (CELP) and Multi-Band Excitation coding (MBE) are typical. For each technology, see "Code
-excited linear prediction (CELP): High-qualit
y speech at 8kbps ”(MR Shroeder and BSAtal,
ICASSP'85, pp.937-940,1985), and "A real-time im
plementation of the improved MBE speech coder "
(MS Brandstein, PAMonta, JCHardwick and JS
Lim, ICASSP'90, pp.5-8, 1990).

Here, CELP audio coding will be described. FIG. 12 shows an example of the overall configuration of a CELP speech coding / decoding method.
Is a decoding unit, 3 is a multiplexing unit, and 4 is a demultiplexing unit. The coding unit 1 includes a linear prediction parameter analysis unit 8, a linear prediction parameter coding unit 9, a synthesis filter 10, an adaptive codebook 11, a noise codebook 12, a gain coding unit 13, and a distance calculation unit 14. Further, the decoding unit 2 includes a linear prediction parameter decoding unit 15, a synthesis filter 16,
Adaptive codebook 17, random codebook 18, gain decoding means 1
It is composed of 9.

In CELP speech coding, one frame is about 5 to 50 ms, and the speech of the frame is divided into spectrum information and sound source information and coded. The operation of the CELP speech coding / decoding method will be described below. First, in the encoding unit 1, the linear prediction parameter analysis unit 8 analyzes the input speech S1 and extracts a linear prediction parameter which is spectrum information of the speech. The linear prediction parameter coding means 9 codes the linear prediction parameter and sets the coded linear prediction parameter as a coefficient of the synthesis filter 10.

Next, encoding of sound source information will be described.
The past drive excitation vector is stored in the adaptive codebook 11, and the time series vector obtained by periodically repeating the past drive excitation vector corresponding to the adaptive code is output. The random codebook 12 stores a plurality of time series vectors generated from random noise, for example, and outputs the time series vector corresponding to the random code. Each time series vector from the adaptive codebook 11 and the noise codebook 12 is gain coding means 1.
Weights are added according to the respective gains given by 3 and added, and the addition result is supplied to the synthesis filter 10 as a driving excitation vector to obtain encoded speech. The distance calculation means 14 obtains the distance between the encoded voice and the input voice S1, and searches for an adaptive code, a noise code, and a gain that minimize the distance. After this coding is completed, the code of the linear prediction parameter, the adaptive code that minimizes the distortion between the input speech and the coded speech, the noise code, and the gain code are output as the coding results.

On the other hand, in the decoding unit 2, the linear prediction parameter decoding means 15 decodes the linear prediction parameter from the code of the linear prediction parameter and sets it as the coefficient of the synthesis filter 16. Next, adaptive codebook 17 outputs a time-series vector obtained by periodically repeating the past driving excitation vector corresponding to the adaptive code, and noise codebook 18 outputs a time-series vector corresponding to the noise code. To do. These time-series vectors are weighted and added according to the respective gains decoded from the gain code by the gain decoding means 19, and the addition result is supplied to the synthesis filter 16 as a driving sound source vector to output the output voice S3. can get.

Further, as a conventional speech encoding / decoding method improved for the purpose of improving the reproduced speech quality by the CELP system speech encoding / decoding method, there is one disclosed in Japanese Patent Laid-Open No. 6-202699. FIG. 13 in which parts corresponding to those in FIG. 12 are assigned the same reference numerals shows the configuration of this conventional speech coding / decoding method, in which 20 and 22 are pitch position extracting means.
Reference numerals 21 and 23 are pitch synchronization means. The operation of the encoding / decoding method having such a configuration will be described. First, in the encoding unit 1, the pitch position extraction means 20 extracts, from the periodically repeated time series vector output from the adaptive codebook 11, for example, a periodic point having the maximum amplitude of the time series vector as a pitch position. To do. In the noise codebook 12,
For example, a plurality of code vectors generated from random noise are stored, and a code vector corresponding to a noise code is output.

A pitch synchronizing position is set for each code vector, and the pitch synchronizing means 21 cuts out the code vector so that the pitch synchronizing position matches the pitch position extracted by the pitch position extracting means 20. A period length is set and a time series vector is generated by repeating this periodically. The method of generating this time series vector is, for example,
"Study on Phase-Adaptive PSI-CELP Speech Coding" (Mano, Moriya, IEICE Speech Society SP94-96, p.
p.37-44, 1995), a code vector is cut out according to two parameters of a frame boundary, a pitch position shift (phase), and a pitch cycle length, and the code vector is made periodic. FIG. 14 shows an example of pitch synchronization processing when the phase is φ and the pitch period is L. Then, the coded speech is generated using each time-series vector from the adaptive codebook 11 and the pitch synchronization means 21, and the adaptive code, noise code, and gain that minimize the distance between the coded speech and the input speech S1. Is selected and encoded.

Next, in the decoding unit 2, the pitch position extracting means 22 uses the same method as the pitch position extracting means 20 of the encoding unit 1 from the periodically repeated time series vector output from the adaptive codebook 17. Extract the pitch position. The random codebook 18 also outputs a code vector corresponding to the random code. For the code vector, the random codebook 1 of the encoding unit 1 is used.
The same pitch synchronization position as that of No. 2 is set, and the pitch synchronizing means 23 cuts out the code vector to make the pitch period length so that the pitch synchronizing position matches the pitch position extracted by the pitch position extracting means 22. To periodically generate a time series vector.

Then, the output speech S3 is obtained by using each time series vector from the adaptive codebook 17 and the pitch synchronizing means 23. With this configuration, it is possible to improve the pitch periodicity of the reproduced voice by synchronizing the code vector from the random codebook with the pitch, and particularly to improve the quality of the reproduced voice in the voiced part of the voice, and Since the pitch position when the code vector is made periodic is adaptively obtained from the time-series vector from the adaptive codebook, high-quality reproduced speech can be generated without increasing the amount of transmission information.

Here, in a speech coding / decoding method which is actually applied to an application field in which a coding error occurs, such as mobile communication, an error correction coding technique is used to deteriorate the quality of coded speech due to a coding error. Holding down. Further, when the error cannot be completely corrected, a waveform restoration process is performed to reduce the influence of the code error.

There are the following two types of repair methods so far. In other words, the first repair method is "Channel codi
ng for digital speech transmission in the Japanese
digital cellular system ”(MJMcLaughlin, Radio Communication System Research Society of the Institute of Electronics, Information and Communication Engineers RCS90-27, pp.
41-45, 1990), when the current frame is a frame with a code error, the reproduced voice is generated by repeatedly using the parameters of the past frame, and the power of the reproduced voice is gradually suppressed. That's the way to go.

The second repair method is disclosed in Japanese Patent Laid-Open No. 6-12.
As disclosed in Japanese Patent Publication No. 095, the code error detection information of each of the past frame, the current frame, and the future frame is used to reproduce and restore the voice of the current frame according to each frame error detection state. Is. In this case, since the interpolation is also performed using the information of the future frame, it is possible to perform the interpolation with less distortion as compared with the case where only the information of the past frame is used.

Also, when the waveform restoration cannot be performed sufficiently,
As a transmission error compensation method that maintains the perceptual voice quality,
There is one disclosed in JP-A-7-36496.
In this, the reproduced voice is not output due to a transmission error state, but a noise signal is output instead. As a result, it is possible to avoid making an abnormal noise even when proper waveform restoration cannot be performed.

[0015]

In the conventional speech coding / decoding method improved as described above, the code vector in the noise codebook corresponding to the noise code is converted into the time series vector output from the adaptive codebook. The cut-out pitch period length is set so that the pitch synchronization position of the code vector matches the more obtained pitch position, and a time-series vector is generated by repeating this periodically, and this is used to generate the driving sound source vector. When a transmission error occurs in the noise code, the driving excitation vector is affected by the error, and the time series vector from the adaptive codebook generated from this driving excitation vector is also affected by the error. Therefore, the pitch position cannot be obtained correctly.

For this reason, once the noise code is erroneous, even if the correct noise code is transmitted after that and the correct noise code vector is used, the pitch position is different from the original correct one, so that it is generated by the pitch synchronization processing. The correct time series vector cannot be obtained, and the driving sound source vector cannot be correctly generated. Since this causes a vicious circle that causes the pitch position to not be obtained correctly, there is a problem that the influence of the error of the noise code has a large time ripple and the reproduced voice is greatly deteriorated when the noise code is wrong. . Even if the conventional error correction coding technology is used to suppress the quality deterioration of coded speech due to a code error, this noise code must be excluded from error correction in order to reduce the amount of transmission information related to the error correction code. Therefore, it is necessary to suppress the quality deterioration due to the noise code error as small as possible.

In the pitch synchronization processing, when the code vector in the random codebook corresponding to the random code is cut out to be the pitch cycle length, two parameters of the phase and the pitch cycle length are used. If the pitch position is incorrect and the phase is different from the original correct one, the waveform represented by the code vector of the pitch cycle length is different from the original correct one, and the time-series vector waveform that repeats this periodically is also the original correct one. There is a problem that the reproduced voice is greatly deteriorated because it is very different from the one. FIGS. 15A and 15B show examples of time-series vectors generated by the pitch synchronization processing when the code vector has the same pitch period length but only the pitch position is different.

Further, in the conventional waveform restoration method when a code error occurs, the parameters of the past frame are repeatedly used, but there is a problem that the quality of reproduced voice is still low. For example, in a voiced rising frame, the pitch periodicity is non-stationary, so the pitch information parameter suitable for the subsequent voiced stationary part is not always transmitted, and when a code error occurs in the subsequent frame, Even if the parameters of pitch information were used repeatedly, good reproduced speech could not be obtained.

Further, even in the unvoiced part, there is a frame having local periodicity, and when a code error occurs in the unvoiced frame following the frame, the signal of a constant period continues due to the repetition of the parameter, so that the reproduced voice is a buzzer sound. Therefore, the quality of the reproduced voice was greatly deteriorated. Further, in a frame in which a code error has occurred, if the spectrum generated by repeating the parameters of the previous frame is inappropriate, the reproduced voice quality is greatly deteriorated. This is especially true when there is a large amount of power in the high range, such as a sharp formant peak in the high range.
The abnormal noise in the high frequency range became noticeable, and the auditory deterioration was great.

Further, the conventional waveform restoration method for performing the interpolation by using the information of the future frame also has a problem that the delay required for reproducing the current frame for restoration becomes large. If the delay required for this voice reproduction becomes large, when it is applied to communication, a natural dialogue cannot be performed and a telephone call is disturbed. Therefore, it is desirable that the delay be as small as possible. Further, even when a conventional code error occurs and waveform restoration cannot be performed sufficiently, the transmission error compensation method for maintaining the audible voice quality switches and outputs the reproduced voice and the noise signal.
However, since a discontinuity is involved in the transition from voice to noise or from noise to voice, there is a problem that this rather leads to deterioration of the reproduced voice quality in terms of hearing. Furthermore, when a transmission error occurs, a noise signal is always output, so noise is output even in the originally silent portion, which also leads to deterioration in hearing.

The present invention has been made to solve the above problems, and provides a speech coding method, a speech decoding method, and a speech coding / decoding method for reproducing speech with less quality deterioration due to a transmission path error. It is a thing. The present invention also provides a voice decoding method and a voice code which, when a code error occurs, satisfactorily reproduce and restore the voice of the current frame without increasing the delay and reproduce the voice with less quality deterioration due to a transmission path error. An encryption / decryption method is provided.
Furthermore, the present invention provides a speech decoding method and a speech encoding / decoding method that superimpose noise on reproduced speech when a code error occurs to maintain audible speech quality.

[0022]

[Means for Solving the Problems ]
For this reason, the speech encoding method of the present invention is based on
Memory, and corresponding to the adaptive code, past drive sound source
Outputs the first time-series vector in which Toll is repeated periodically
With an adaptive codebook and a predetermined pitch synchronization position
The code corresponding to the noise code is stored by storing multiple code vectors
And a random codebook for outputting a signal vector,
Based on any of the signs of the above parameters except
From the characteristic positions lined up at pitch intervals in the frame,
The pitch position is extracted, and each of the above-mentioned pitch positions is extracted.
Adjust the pitch circumference so that the pitch synchronization position of the code vector matches.
Second time when the code vector with period length is repeated cyclically
Generate a sequence vector and add it to the first time-series vector
The driving time source vector is calculated by adding the second time series vector.
Generate and drive the synthesis filter with the above driving sound source vector
Input audio of the reproduced audio signal
The distortion is evaluated to determine the sign.

The speech decoding method of the next invention is the same as the speech decoding method of the past.
The driving sound source vector is stored, and the past
The first time-series vector in which the driving sound source vector is periodically repeated.
An adaptive codebook that outputs a cutout and a predetermined pitch
Stores multiple code vectors with synchronization positions
Equipped with a random codebook that outputs a code vector corresponding to
Any of the codes of the plurality of parameters excluding the noise code
Based on the
Pitch position is extracted from the position, and the extracted pitch is
Match the pitch synchronization position of each code vector above to the position
, The code vector with the pitch cycle length is periodically repeated.
To generate the second time series vector
Driving sound by adding the vector and the second time series vector
Generate a source vector and combine it with the above driving source vector.
Drive the audio output.

A speech encoding / decoding method of the next invention is
On the encoding side, the past driving excitation vector is stored and adapted.
The past driving sound source vector is periodically repeated corresponding to the code.
An adaptive codebook that outputs the returned first time series vector,
A code vector with a predetermined pitch synchronization position
Stores multiple codes and outputs a code vector corresponding to the noise code
A random codebook that
Based on one of the parameter signs,
Pitch positions are extracted from the characteristic positions lined up at pitch intervals,
The pitch of each code vector is added to the extracted pitch position.
The pitch is set so that the sync position is matched.
Generates a second time-series vector that repeats Toll periodically
The first time series vector and the second time series vector
Drive sound source vector is generated by adding the torque
Playback audio signal by driving synthesis filter with source vector
And evaluate the distortion of the reproduced voice signal with respect to the input voice.
To determine the code, and the decoding side
Memory, and corresponding to the adaptive code, past drive sound source
Outputs the first time-series vector in which Toll is repeated periodically
With an adaptive codebook and a predetermined pitch synchronization position
The code corresponding to the noise code is stored by storing multiple code vectors
And a random codebook for outputting a signal vector,
Based on any of the signs of the above parameters except
From the characteristic positions lined up at pitch intervals in the frame,
The pitch position is extracted, and each of the above-mentioned pitch positions is extracted.
Adjust the pitch circumference so that the pitch synchronization position of the code vector matches.
Second time when the code vector with period length is repeated cyclically
Generate a sequence vector and add it to the first time-series vector
The second time series vector is weighted and added.
Drive source vector is generated by
A synthesis filter is driven to generate an output voice.

The speech encoding method of the next invention is the same as that of the past.
The driving sound source vector is stored, and the past
The first time-series vector in which the driving sound source vector is periodically repeated.
An adaptive codebook that outputs a cutout and a predetermined pitch
The synchronization position and the preset cutoff corresponding to the pitch cycle.
Stores a plurality of code vectors with output positions,
Equipped with a random codebook that outputs a code vector corresponding to
From the characteristic positions lined up at pitch intervals in the current frame
Pitch position is extracted and each of the above code vectors is pitch period
Cut out at the cutout position corresponding to
The pitch synchronization position of the signal vector is the pitch position extracted above.
The second time-series vector that is periodically repeated to match the position
To generate the first time series vector and the second time series
Driving sound source by weighting and adding sequence vectors
Generate a vector and synthesize fill with the above driving sound source vector
Drive the audio signal to reproduce the audio signal, and the reproduced audio signal above.
The code is determined by evaluating the distortion of the input voice of.

Further, in the speech coding method of the next invention, the frame has an integer multiple of the pitch period.

Further, in the speech coding method of the next invention, a fixed length which is a standard of a frame is determined, and an integer multiple of the pitch period is set to be closest to the fixed length.

Further, in the speech encoding method of the next invention, a fixed length which is a standard of the frame is determined, and the average of integral multiples of the pitch period is always equal to or larger than the fixed length.

Further, in the speech encoding method of the next invention, a fixed point serving as a standard of a frame is determined, and an integer multiple of the pitch period is set so that the frame boundary is closest to the fixed point.

The speech decoding method of the next invention is the same as that of the past.
The driving sound source vector is stored, and the past
The first time-series vector in which the driving sound source vector is periodically repeated.
An adaptive codebook that outputs a cutout and a predetermined pitch
The synchronization position and the preset cutoff corresponding to the pitch cycle.
Stores a plurality of code vectors with output positions,
Equipped with a random codebook that outputs a code vector corresponding to
From the characteristic positions lined up at pitch intervals in the current frame
The pitch position is extracted and the above code vectors are set to the pitch period.
Cut out at the corresponding cutout position, and this cut out code
The pitch sync position of the vector is the pitch position extracted above
Second time series vector that is periodically repeated to match
To generate the first time series vector and the second time series
Each column vector is weighted and added to add the driving sound source vector.
Generate a cuttle and synthesize filter with the above driving sound source vector
To produce output audio.

Further, in the speech decoding method of the next invention, the frame has an integer multiple of the pitch period.

Further, in the speech decoding method of the next invention, a fixed length which is a standard of a frame is determined, and an integer multiple of the pitch period is set to be closest to the fixed length.

Further, in the speech decoding method of the next invention, a fixed length which is the standard of the frame is set, and the average of integral multiples of the pitch period is always equal to or larger than the fixed length.

Further, in the speech decoding method of the next invention, a fixed point serving as a standard of a frame is determined, and an integer multiple of the pitch period is set so that the frame boundary is closest to the fixed point.

A speech encoding / decoding method of the next invention is
On the encoding side, the past driving excitation vector is stored and adapted.
The past driving sound source vector is periodically repeated corresponding to the code.
An adaptive codebook that outputs the returned first time series vector,
The preset pitch synchronization position and the pitch period
The code vector with the cutout position determined accordingly.
Stores multiple codes and outputs a code vector corresponding to the noise code
With a random codebook for
The pitch positions are extracted from the characteristic positions arranged at intervals and
Cut out the vector at the cutting position corresponding to the pitch period
However, the pitch synchronization position of this clipped code vector is
Periodically repeated to match the extracted pitch position
Second time series vector is generated, and the first time series vector is generated.
The weight of Khuttle and the second time series vector
And add to generate a driving sound source vector
Drive the synthesis filter with a vector to reproduce the audio signal,
Evaluate the distortion of the reproduced voice signal with respect to the input voice
The code is decided, and on the decoding side, the past driving sound source vector
And the past driving sound source vector corresponding to the adaptive code
Outputs the first time-series vector that is repeated periodically
Adaptive codebook, predetermined pitch synchronization position, and
Has a cutout position that is determined according to the pitch period
Multiple code vectors are stored and the code vector corresponding to the noise code is stored.
Equipped with a random codebook that outputs a cutout,
Pitch positions are extracted from the characteristic positions lined up at pitch cycle intervals.
Then, cut out each of the above code vectors corresponding to the pitch period.
Cut out at this position, and
The pitch is adjusted so that the synchronization position matches the pitch position extracted above.
Generate a second time series vector that is repeated periodically,
The first time series vector and the second time series vector are
Generate driving sound source vectors by weighting and adding each
Then, drive the synthesis filter with the above driving sound source vector and output.
Generate force voice

The speech coding apparatus of the next invention is the same as the speech coding apparatus of the past.
The driving sound source vector is stored, and the past
The first time-series vector in which the driving sound source vector is periodically repeated.
An adaptive codebook that outputs a cutout and a predetermined pitch
Stores multiple code vectors with synchronization positions
A random codebook that outputs a code vector corresponding to
Select any of the codes of the above multiple parameters excluding the noise code
Based on the feature position of the current frame
Pitch position extraction means for extracting the pitch position from the
The pitch of each of the above code vectors at the extracted pitch position
Code vector with pitch period length so that the synchronization positions match
Generate a second time series vector that repeats
Pitch synchronization means and the first time-series vector
The driving time source vector is calculated by adding the second time series vector.
Addition means to generate and audio signal with the above driving sound source vector
And the input of the reproduced audio signal.
Distance calculator that evaluates distortion for force speech and determines the code
And a step.

The speech decoding apparatus of the next invention is the same as the speech decoding apparatus of the past.
The driving sound source vector is stored, and the past
The first time-series vector in which the driving sound source vector is periodically repeated.
An adaptive codebook that outputs a cutout and a predetermined pitch
Stores multiple code vectors with synchronization positions
A random codebook that outputs a code vector corresponding to
Select any of the codes of the above multiple parameters excluding the noise code
Based on the feature position of the current frame
Pitch position extraction means for extracting the pitch position from the
The pitch of each of the above code vectors at the extracted pitch position
Code vector with pitch period length so that the synchronization positions match
Generate a second time series vector that repeats
Pitch synchronization means and the first time-series vector
The driving time source vector is calculated by adding the second time series vector.
The adding means for generating and the synthesizing filter for generating the output sound by the driving sound source vector are provided.

The speech coding apparatus of the next invention is the same as the speech coding apparatus of the past.
The driving sound source vector is stored, and the past
The first time-series vector in which the driving sound source vector is periodically repeated.
An adaptive codebook that outputs a cutout and a predetermined pitch
The synchronization position and the preset cutoff corresponding to the pitch cycle.
Stores a plurality of code vectors with output positions,
The random codebook that outputs the code vector corresponding to
From the characteristic positions lined up at pitch period intervals in the rem
Pitch position extracting means for extracting the H position, and
Cut out the cuttle at the cutout position corresponding to the pitch cycle.
However, the pitch synchronization position of this clipped code vector is
Periodically repeated to match the extracted pitch position
Pitch synchronization means for generating a second time-series vector
And the first time series vector and the second time series vector
Adding means for adding driving torque to generate a driving sound source vector
And a synthetic stream that reproduces an audio signal with the above driving sound source vector.
And the distortion of the reproduced audio signal with respect to the input audio.
And a distance calculating means for determining the sign.

The speech decoding apparatus of the next invention is the same as the speech decoding apparatus of the past.
The driving sound source vector is stored, and the past
The first time-series vector in which the driving sound source vector is periodically repeated.
An adaptive codebook that outputs a cutout and a predetermined pitch
The synchronization position and the preset cutoff corresponding to the pitch cycle.
Stores a plurality of code vectors with output positions,
The random codebook that outputs the code vector corresponding to
From the characteristic positions lined up at pitch period intervals in the rem
Pitch position extracting means for extracting the H position, and
Cut out the cuttle at the cutout position corresponding to the pitch cycle.
However, the pitch synchronization position of this clipped code vector is
Periodically repeated to match the extracted pitch position
Pitch synchronization means for generating a second time-series vector
And the first time series vector and the second time series vector
Driving sound source vector
And the output sound from the driving sound source vector
And a synthesis filter for generating a voice.

[0040]

[0041]

[0042]

[0043]

[0044]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

Embodiment 1. FIG. 1 in which parts corresponding to those in FIG. 13 are assigned the same reference numerals shows the overall configuration of the first embodiment of the speech coding method and speech decoding method according to the present invention. A fixed codebook used by switching to the codebook 11, 25 is a second random codebook used together with the fixed codebook 24, and 26 is a pitch position extraction means. In the decoding unit 2 in the figure, 27 is a fixed codebook used by switching from the adaptive codebook 17, 28 is a second random codebook used together with the fixed codebook 27, and 29 is a pitch position extraction means.

The operation will be described below. First, the encoding unit 1
In the linear prediction parameter analysis means 8, the input speech S
1 is analyzed, and a linear prediction parameter that is the spectrum information of the voice is extracted. The linear prediction parameter coding means 9 codes the linear prediction parameter and sets the coded linear prediction parameter as a coefficient of the synthesis filter 10. Here, encoding of sound source information will be described. In order to code the excitation information to generate the driving excitation vector, two cases are used: the case of using the adaptive codebook 11 and the random codebook 12, and the case of using the fixed codebook 24 and the second random codebook 25.
Assuming that a method for generating a driving sound source vector in which the distance between the coded speech and the input speech becomes smaller is selected, these two methods are switched and used. Here, the case of using the fixed codebook 24 and the second random codebook 25 will be described first, and then the adaptive codebook 1 will be described.
The case of using 1 and the random codebook 12 will be described.

The fixed codebook 24 stores a plurality of time series vectors generated from random noise, for example.
The time series vector corresponding to the adaptive code is output. The second random codebook 25 stores, for example, a plurality of time series vectors generated from random noise, and outputs a time series vector corresponding to the random code. The time series vectors from the fixed codebook 24 and the second random codebook 25 are weighted and added according to the respective gains given from the gain coding unit 13, and the addition result is used as a driving excitation vector to synthesize the filter 10. To obtain encoded speech.
The distance calculation means 14 obtains the distance between the encoded voice and the input voice S1, and searches for an adaptive code, a noise code, and a gain that minimize the distance.

Next, the case where the adaptive codebook 11 and the random codebook 12 are used will be described. In the adaptive codebook 11,
The past driving sound source vector is stored, and a time series vector obtained by periodically repeating the past driving sound source vector corresponding to the adaptive code is output. Pitch position extraction means 26
If the fixed codebook is selected in the previous frame, corresponds to the adaptive code searched in the current frame from the time series vector output from the fixed codebook 24 in the previous frame, for example, at the end of the time series vector. The point obtained in the cycle corresponding to the adaptive code is extracted as the pitch position with reference to the point having the maximum amplitude in the range of the cycle length.

When the adaptive codebook is selected in the previous frame, a point repeated at a cycle corresponding to the adaptive code searched in the current frame is extracted as a pitch position with reference to the pitch position obtained in the previous frame. To do. Figure 2
An example of the pitch position extraction process when the period corresponding to the adaptive code searched in the current frame is L in the first embodiment is shown in FIG. FIG. 2A shows an example when the fixed codebook is selected in the previous frame, and FIG. 2B shows an example when the adaptive codebook is selected in the previous frame.

The random codebook 12 stores a plurality of code vectors generated from random noise, for example, and outputs a code vector corresponding to a random code. A pitch synchronization position is set for each code vector, and the pitch synchronization means 21 cuts out the code vector and sets the pitch cycle length so that the pitch synchronization position matches the pitch position extracted by the pitch position extraction means 26. , Which is repeated periodically to generate a time series vector. The time-series vectors from the adaptive codebook 11 and the pitch synchronization means 21 are weighted and added according to the respective gains given from the gain coding unit 13, and the addition result is sent to the synthesis filter 10 as a driving excitation vector. The supplied coded speech is obtained. The distance calculation means 14 obtains the distance between the encoded voice and the input voice S1, and searches for an adaptive code, a noise code, and a gain that minimize the distance.

When the adaptive codebook 11 and the random codebook 12 are used, the fixed codebook 24 and the second random codebook 2 are used.
5 and the case where 5 is used, the distance between the coded speech generated using the adaptive code, the noise code, and the gain selected respectively and the input speech S1 is compared, and the adaptive code and the noise with the smaller distance are compared. The sign and gain are selected. After the above coding is completed, the code of the linear prediction parameter, the adaptive code that minimizes the distortion between the input speech and the coded speech, the noise code,
The sign of the gain is output as the coding result S2. The above is the characteristic operation of the speech coding method according to the first embodiment.

Next, the decoding unit 2 will be described. In the decoding unit 2, the linear prediction parameter decoding unit 15 decodes the linear prediction parameter from the code of the linear prediction parameter and sets it as the coefficient of the synthesis filter 16. Here, decoding of sound source information will be described. There are two methods for decoding the excitation information and generating the driving excitation vector: using adaptive codebook 17 and random codebook 18, and using fixed codebook 27 and second random codebook 28. Assuming that a method is used and the method used for encoding is selected corresponding to the adaptive code and the noise code, these two methods are switched and used.

Here, the case of using the fixed codebook 27 and the second random codebook 28 will be described first, and then the case of using the adaptive codebook 17 and the random codebook 18 will be described. When the fixed codebook 27 and the second random codebook 28 are used, the fixed codebook 27 outputs the time series vector corresponding to the adaptive code, and the second random codebook 28 outputs the time series vector corresponding to the random code. Output the series vector. Next, the case of using the adaptive codebook 17 and the random codebook 18 will be described. The adaptive codebook 17 outputs a time-series vector obtained by periodically repeating the past drive excitation vector corresponding to the adaptive code. When the fixed codebook is selected in the previous frame, the pitch position extraction means 29 selects the adaptive codebook in the previous frame from the time series vector output from the fixed codebook 27 in the previous frame. In this case, the pitch position is extracted from the pitch position obtained in the previous frame by the same method as the pitch position extracting means 26 of the encoding unit 1.

The random codebook 18 also outputs a code vector corresponding to the random code. A pitch synchronization position is set in the code vector, and the pitch synchronization means 23 cuts out the code vector so that the pitch synchronization position matches the pitch position extracted by the pitch position extraction means 29, This is periodically repeated to generate a time series vector. Fixed codebook 27 and second random codebook 2
The time series vector from 8 or the time series vector from the adaptive codebook 17 and the pitch synchronizing means 23 is weighted and added according to each gain decoded from the gain code by the gain decoding means 19, The addition result is supplied to the synthesis filter 16 as a driving sound source vector, and the output voice S3 is obtained. The above is the operation characteristic of the speech decoding method according to the first embodiment.

According to the first embodiment, the pitch position at the time of pitch synchronization of the noise code vector is obtained from the information related to the adaptive code only, and is determined regardless of the information related to the noise code. Even if a transmission error occurs in the code, if the adaptive code is transmitted correctly, the pitch position is not affected by the error, so the influence of the noise code error is small in time and avoids deterioration of the reproduced voice. ,
It is possible to reproduce voice with little quality deterioration due to transmission error.

Embodiment 2. Pitch synchronization processing in the second embodiment of the speech coding method and speech decoding method according to the present invention will be described. Each code vector in the random codebook is set with a position at which the code vector is cut out corresponding to the pitch cycle, and the code vector is cut out from this position to have a pitch cycle length. Then, a time series vector is generated by repeating this periodically so that the pitch synchronization position matches the pitch position. FIG. 3 shows an example of the time-series vector generated by the pitch synchronization processing according to the second embodiment. When the pitch cycle is L in the figure, the position where the code vector is cut out is φ (L) before the pitch synchronization position, and φ
(L) is determined from only the pitch period, for example, L / 2. 3A and 3B, the code vector and the pitch period length are the same, only the pitch position is different, and the waveforms represented by the pitch period code vector are the same in FIGS. 3A and 3B. A time-series vector obtained by periodically repeating this is also the same waveform shifted in time.

According to the second embodiment, when the code vector in the random codebook corresponding to the random code is cut out in the pitch synchronization process to be the pitch cycle length, only the pitch cycle length is used as a parameter and the pitch position is set. , The waveform represented by the code vector of the pitch cycle length is the same as the original correct one, even if the pitch position is incorrect. Since the correct one is temporally shifted, the auditory deterioration due to the error in the pitch position is small, the deterioration of the reproduced sound can be avoided, and the sound with less quality deterioration due to the transmission error can be reproduced.

Third Embodiment In the second embodiment described above,
The position where the code vector is cut out in the random codebook is φ (L) before the pitch synchronization position, and φ (L) is determined by using the same function regardless of the pitch cycle, such as L / 2. Instead of this, the function that determines the position where the code vector is cut out for each pitch period may be changed. Further, the function for determining the position where the code vector is cut out may be changed for each pitch period for each code vector in the random codebook. According to the third embodiment, the position where the code vector is cut out can be freely set for each pitch period or each code vector. Therefore, by learning and acquiring this setting so as to improve the quality of the reproduced voice, the transmission is performed. Even if there is no error, the quality of synthesized speech can be improved.

Fourth Embodiment In the above-described Embodiments 2 and 3, the code vector in the same random codebook is used regardless of the pitch cycle, but instead of this, a plurality of random codebooks are provided and the random code is used according to the pitch cycle. You may switch books. According to the fourth embodiment, different random codebooks can be used depending on the pitch period, so that the degree of freedom of the time series vector that can be generated is increased as compared with the case where one conventional random codebook is used. Even if there is no transmission error, the quality of synthesized speech can be improved.

Fifth Embodiment In Embodiment 4 described above, a random codebook different for every pitch cycle length is further provided,
The random codebook corresponding to each pitch cycle length may be composed of code vectors of that pitch cycle length. According to the fifth embodiment, since the code vector in the random codebook corresponding to each pitch cycle is already the pitch cycle length, the code vector is cut out in the pitch synchronization processing and the processing for setting the pitch cycle length is omitted. You can

Sixth Embodiment In the second embodiment described above,
Although the frame, which is a unit for generating the driving sound source vector, has a fixed length, the frame may have a variable length according to the pitch cycle length, and the frame may have an integral multiple of the pitch cycle. FIG. 4 shows an example of a time-series vector generated by the pitch synchronization processing of the noise code vector when the frame in this Embodiment 6 has an integral multiple of the pitch. In FIGS. 4A and 4B, the code vector and the pitch period length are the same, only the pitch position is different, the generated time-series vectors are the same waveform, and the difference in the pitch position is the frame boundary. Appears to be shifted in time.

According to the sixth embodiment, the frame, which is a unit for generating the driving sound source vector, has a variable length according to the pitch cycle length, and the frame has an integral multiple of the pitch cycle. If the position is incorrect, the problem that the waveform of the time series vector generated from the random codebook at the frame boundary part was different from the original correct one can be solved, and the time series vector from the generated random codebook can be solved. Has the exact same waveform as the original one, and the effect of pitch position error is that the frame only shifts temporally, so the auditory deterioration due to the pitch position error is small, and the deterioration of the reproduced voice is avoided. However, it is possible to reproduce voice with less quality deterioration due to transmission error.

Embodiment 7. Further, in the above-described sixth embodiment, a certain fixed length is defined as a frame standard, and a variable length frame that is an integral multiple of the pitch period is determined to have an integer multiple value so as to be closest to the fixed length. Also good. FIG. 5 shows an example of a frame when the fixed length serving as the reference of the frame period is N and the pitch period is L. According to the seventh embodiment, the frame length which is a unit for encoding is maintained at a constant value to some extent, so that the encoded information amount (bit rate) in a unit time can be stabilized and the same time length can be obtained. In the case of voice, it can be encoded with almost the same amount of information regardless of the pitch cycle length.

Embodiment 8. Further, in the above-described sixth embodiment, a certain fixed length is set as a frame standard, and a value that is an integral multiple is determined so that the average of a variable length frame that is an integral multiple of the pitch period is always the fixed length or more. Also good. FIG. 6 shows an example of a frame where N is a fixed length and L is a pitch period, which are criteria of the frame period. According to the eighth embodiment, the time length of voice coded with a variable length frame of a certain number of frames is always equal to or longer than the time length of voice coded with a fixed length frame of the same number of frames. Therefore, when applied to communication, the delay on the decoding side can be reduced. In addition, since the average of the period of the frame, which is the unit for encoding, is always kept above a certain value, it is possible to specify the maximum value of the encoded information amount (bit rate) in a unit time, and the audio of the same time length In this case, it is possible to perform encoding with an amount of information equal to or less than the information amount determined by the maximum value of the defined bit rate regardless of the pitch cycle length.

Ninth Embodiment Furthermore, in the sixth embodiment described above, the variable-length frame that is an integral multiple of the pitch period is
It is also possible to set a certain fixed point as a criterion of the frame cycle and determine a value that is an integral multiple so that the frame boundary is closest to the fixed point. FIG. 7 shows an example of a frame in which a fixed point serving as a frame cycle criterion is a point for each cycle N and the pitch cycle is L. According to the eighth embodiment, the period of a frame, which is a unit for encoding, is maintained at a constant value to some extent, so that the encoded information amount (bit rate) in a unit time can be stabilized and the same time length can be obtained. The voice can be encoded with almost the same amount of information regardless of the pitch cycle length.

Tenth Embodiment Embodiments 6 to 9 described above
However, since the time information indicating at what point in time the input and encoded frames are encoded and decoded is not transmitted at all, on the decoding side, the decoded frame corresponds to what point in the input speech. Can be determined only by integrating the variable frame length from the time when decoding is started to the frame. Therefore, even if the frame length is erroneously decoded due to a transmission error even once, the reproduced voice generated on the decoding side thereafter has a time lag with respect to the input voice, and this lag cannot be eliminated. was there. Therefore, instead of transmitting no time information at all, the time information may be transmitted at least intermittently, for example, by transmitting a time difference from the coding start time at the time of coding the frame.

According to the tenth embodiment, since the time information is transmitted at least intermittently, the reproduced voice generated by decoding the frame length erroneously due to a transmission error and the decoding side is temporally different from the input voice. Even if a shift occurs, it is possible to compensate for and eliminate the time shift by using the transmitted time information, so that it is possible to reproduce voice with less quality deterioration due to a transmission error.

Eleventh Embodiment FIG. 8 in which the same parts as those in FIG. 1 are assigned the same reference numerals shows the configuration of the eleventh embodiment of the speech decoding method according to the present invention, in which 30 is a linear prediction parameter when a transmission error of the code is detected. A linear prediction parameter repairing means for repairing, a rising edge determining means for analyzing the reproduced speech to determine whether or not the preceding frame is a voiced rising edge, a pitch extracting means for analyzing the reproduced speech to obtain a pitch, and a reference numeral 33 Is an adaptive code repairing means for repairing the adaptive code when the transmission error is detected. Reference numeral 34 is an unvoiced determination means for analyzing the reproduced voice to determine whether or not the previous frame is unvoiced, 35 is a gain recovery means for recovering a gain when a code transmission error is detected, and 36 is a code transmission error detection. It is a reproduced voice power control means for controlling the power of the reproduced voice according to the situation. Further, 37 is a voice state estimating means for estimating the voice state of the current frame when a code transmission error is detected, 38 is noise generating means for generating a noise signal, and 39 is noise depending on a code transmission error detection situation. It is a noise power control means for controlling the power of the signal.

In the decoding unit 2, the linear prediction parameter decoding means 15 decodes the linear prediction parameter from the code of the linear prediction parameter. The linear prediction parameter repairing means 30 uses the decoded linear prediction parameter in the frame in which no transmission error of the code is detected, and the low band from the linear prediction parameter decoded in the previous frame in the frame in which the transmission error is detected. While maintaining the formant structure of, the linear prediction parameters such as a spectrum in which the high-range formant peak is suppressed, for example, when the linear prediction parameter is expressed as a line spectrum pair parameter, it is determined that only the low-frequency parameters are used, Synthesis filter 1
Set as a coefficient of 6. FIG. 9 shows an example of the spectrum envelope represented by the linear prediction parameter of the previous frame and the spectrum envelope in which the structure of the low band is maintained and the high band is suppressed by the linear prediction parameter restoration process.

The rising edge judging means 31 analyzes the reproduced voice of the previous frame in the frame in which the code transmission error is detected, judges whether or not the preceding frame is a voiced rising edge, and the judgment result is the pitch analyzing means. 32, and outputs to the adaptive code restoration means 33. Here, it is determined whether or not the voiced rising is occurring, for example, when the power of the voice in the second half of the frame in the previous frame is larger than the power of the first half of the frame by a certain threshold value or more. The pitch extracting means 32 is a frame in which a code transmission error is detected, and when the preceding frame is determined to be a voiced rising edge,
The past reproduced speech is analyzed to extract at least one or more pitch period candidates and output to the adaptive code restoration means 33.

The adaptive code restoration means 33 uses the adaptive code input from the separating means 4 in the frame in which no code transmission error is detected, when the preceding frame is not the voiced rising even if a code error is detected. If the code error is detected and the preceding frame has a voiced rising edge, the adaptive code of the preceding frame has been correctly transmitted in the past, for example, from among the pitch period candidates input from the pitch extracting means 32. The code which is closest to the pitch cycle corresponding to the code having a high appearance frequency is selected by comparing with the code appearance frequency distribution, and the adaptive code corresponding to the selected pitch cycle is output to the adaptive codebook 17. The adaptive codebook 17 outputs a time-series vector obtained by periodically repeating the past drive excitation vector corresponding to the adaptive code. FIG. 10A shows an example of adaptive code restoration processing according to the eleventh embodiment, and FIG.
0 (B) shows an example of the restoration process in which the conventional adaptive code of the previous frame is repeatedly used. As shown in the figure, even if the pitch period corresponding to the adaptive code of the previous frame is not suitable for the current frame, good waveform restoration processing can be performed by finding the pitch period suitable for the current frame from the information of the past frames. It turns out that

The random codebook 18 outputs a time series vector corresponding to the random code. The gain decoding means 19 decodes the gain for the time series vector output from the adaptive codebook 17 and the noise codebook 18 from the gain code. The unvoiced determination unit 34 analyzes the reproduced voice of the previous frame in the frame in which the transmission error of the code is detected, determines whether or not the previous frame is unvoiced, and outputs the determination result to the gain restoration unit 35. . The gain restoration means 35 uses the gain input from the gain decoding means 19 in the frame in which no code transmission error is detected, and the gain in the previous frame when the code error is detected and the previous frame is not unvoiced. , If a code error is detected and the previous frame is unvoiced, the gain for the time series vector from the adaptive codebook is multiplied by α and the gain for the time series vector from the noise codebook is β with respect to the gain of the previous frame. Double and output. Here, α and β are, for example, 0 ≦ α <β ≦ 1.

The time-series vectors from the adaptive codebook 17 and the noise codebook 18 are weighted and added according to the respective gains output from the gain restoration means 35, and the addition result is used as a driving excitation vector to synthesize a filter. It is supplied to 16 and reproduced sound is obtained. The reproduced voice power control means 36 suppresses the power of the reproduced voice according to the code transmission error detection situation, for example, by gradually increasing the suppression amount as the error continues. When a code transmission error is detected, the voice state estimating means 37 determines whether the current frame is voiced / non-voiced based on the past reproduced voice and the code of the linear prediction parameter of the current frame, and the determination result is noise generation means. 38
Output to.

The noise generation means 38 detects, for example, an error in accordance with the transmission error detection status of the code and the voiced / non-voiced determination result input from the voice state determination means 37, and estimates the frame as voiced. Generates noise. The noise power control means 39 controls the noise power, for example, by gradually increasing the power at the beginning of noise superimposition and gradually decreasing the power at the end of superimposition. FIG. 11 shows an example of power control processing for reproduced voice and noise signals. The reproduced voice output from the reproduced voice power control means 36 and the noise output from the noise power control means 39 are added together, and the output voice S3 is output.
Is obtained.

According to the eleventh embodiment, when it is detected that the code cannot be correctly decoded, the state of the reproduced voice of the previous frame is judged, and the voice of the current frame is reproduced according to this judgment. By repairing, it is possible to reproduce the voice with less quality deterioration due to the transmission path error without increasing the delay. Further, when a code error occurs, noise can be superimposed on the reproduced voice depending on the state of the reproduced voice, so that the audible voice quality can be maintained.

Twelfth Embodiment In the eleventh embodiment described above, in the linear prediction parameter restoration means 30, when a transmission error occurs, the linear prediction parameter is always obtained from the linear prediction parameter of the previous frame so as to obtain a spectrum in which the formant peak in the high frequency band is suppressed, Synthesis filter 16
However, instead of this, suppression processing is performed only when there is a sharp formant peak in the high band of the spectrum represented by the linear prediction parameter of the previous frame, and in other cases, the linear prediction parameter of the previous frame is used as it is. Alternatively, the processing may be selectively performed according to the state.

According to the twelfth embodiment, when the linear prediction parameter restoration process is performed, the suppression process is performed only when there is a sharp formant peak in the high frequency range and there is a high possibility that the perceived quality will be deteriorated. In other cases, since the linear prediction parameter of the previous frame is repeatedly used, the continuity of the synthesized speech is improved without generating an abnormal sound, and the audible voice quality can be improved.

Thirteenth Embodiment In the eleventh embodiment described above, when the transmission error detection previous frame is unvoiced, the gain restoration means 35 adjusts the gain to suppress the pitch periodicity of the driving sound source vector, but instead of this, For example, the pitch periodicity may be suppressed by using another means such as not using the time series vector from the adaptive codebook. According to the thirteenth embodiment, it suffices to switch the use of the time series vector from the adaptive codebook to suppress the pitch periodicity, and to compare the case where the pitch periodicity is suppressed by adjusting the gain. It can be realized easily.

Fourteenth Embodiment In the eleventh embodiment described above, the voice state estimating means 37 determines whether the current frame is voiced / non-voiced based on the past reproduced voice and the code of the linear prediction parameter of the current frame. The sign of gain may also be used as a parameter for determination. According to the fourteenth embodiment, the presence / absence of sound is determined using not only the past reproduced voice and the code of the linear prediction parameter but also other information, so that the determination can be performed with higher accuracy.

Fifteenth Embodiment In the eleventh embodiment described above, restoration processing at the time of transmission error and noise superposition processing are applied to CELP system speech coding, but instead of this, it is applied to other speech coding systems including MBE system speech coding. Even in this case, similarly, it is possible to improve the audible voice quality of the reproduced voice at the time of transmission error.

[0081]

As described in detail above, according to the inventions of claims 1 to 3 and claims 15 to 16 ,
In the speech coding method and the speech decoding method, since the code used when obtaining the feature position for pitch synchronization of the noise code vector does not include the noise code, the time ripple of the influence of the code error of the noise code is reduced. Therefore, it is possible to avoid deterioration of reproduced sound and reproduce sound with little quality deterioration due to transmission error.

Further, claim 4, claim 9, claim 14 , contract
According to the inventions of claim 17 to claim 18 , in the speech coding method and the speech decoding method, the code vector having the pitch cycle length when pitch-synchronizing the noise code vector is determined only from the pitch cycle. , It is possible to reduce the influence of the error in the pitch position of the time series vector generated from the noise code, avoid the deterioration of the reproduced voice, and reproduce the voice with less quality deterioration due to the transmission error.

According to the fifth and tenth aspects of the present invention, in the speech coding method and speech decoding method, since the frame is an integral multiple of the pitch period, the time series generated from the noise code. It is possible to reduce the influence of the error in the pitch position of the vector, avoid the deterioration of the reproduced sound, and reproduce the sound with less quality deterioration due to the transmission error.

According to the sixth and eleventh aspects of the present invention, in the voice encoding method and the voice decoding method, a fixed length which is a reference of a frame is determined, and a variable length which is an integral multiple of the pitch period. Since the frame of is closest to the fixed length, the encoded information amount (bit rate) per unit time can be stabilized.

According to the seventh and twelfth aspects of the present invention, in the voice encoding method and the voice decoding method, a fixed length serving as a standard of a frame is determined, and a variable length which is an integral multiple of the pitch period. Since the average of the frames is always fixed length or more, the delay on the decoding side can be reduced when applied to communication, and the maximum value of the coded information amount (bit rate) per unit time is specified. can do.

According to the eighth and thirteenth aspects of the present invention, in the voice encoding method and the voice decoding method, a fixed point serving as a standard of a frame is determined, and a frame boundary is closest to the fixed point. Since this is the case, the encoded information amount (bit rate) per unit time can be stabilized.

[0087]

[0088]

[0089]

[0090]

[0091]

[Brief description of drawings]

FIG. 1 is a block diagram showing an overall configuration of a first embodiment of a voice encoding method and a voice decoding method according to the present invention.

FIG. 2 is a schematic diagram for explaining an operation of pitch position extraction processing in the first embodiment of FIG.

[Fig. 3] Fig. 3 is a schematic diagram for explaining pitch synchronization processing of a noise code vector in a second embodiment of a speech coding method and a speech decoding method according to the present invention.

[Fig. 4] Fig. 4 is a schematic diagram for explaining pitch synchronization processing of a noise code vector in a sixth embodiment of a speech coding method and a speech decoding method according to the present invention.

[Fig. 5] Fig. 5 is a schematic diagram for explaining frame determination in Embodiment 7 of a speech encoding method and a speech decoding method according to the present invention.

[Fig. 6] Fig. 6 is a schematic diagram for explaining frame determination in Embodiment 8 of a speech encoding method and a speech decoding method according to the present invention.

[Fig. 7] Fig. 7 is a schematic diagram used for explaining frame determination in Embodiment 9 of a speech encoding method and a speech decoding method according to the present invention.

[Fig. 8] Fig. 8 is a block diagram showing the structure of an eleventh embodiment of a speech decoding method according to the present invention.

FIG. 9 is a schematic diagram for explaining the operation of the spectrum restoration processing in the eleventh embodiment of FIG.

FIG. 10 is a signal waveform diagram for explaining the operation of the adaptive code restoration process in the eleventh embodiment of FIG.

FIG. 11 is a schematic diagram for explaining an example of power control processing of reproduced voice and noise signals in the eleventh embodiment of FIG.

FIG. 12 is a block diagram showing the overall configuration of a conventional CELP-based speech encoding / decoding method.

FIG. 13 is a block diagram showing an overall configuration of a conventional improved CELP audio encoding / decoding method.

FIG. 14 is a schematic diagram for explaining pitch synchronization processing of a noise code vector in the CELP speech coding / decoding method of FIG. 13;

15 is a schematic diagram for explaining pitch synchronization processing of a noise code vector when the phase is incorrect in the CELP speech coding / decoding method of FIG.

[Explanation of symbols]

1 Encoding section 2 Decoding section 3 Multiplexing means 4 Separation means 8 Linear prediction parameter analysis means 9 Linear prediction parameter coding means 10, 16 Synthesis filter 11, 17 Adaptive codebook 12, 18 Noise codebook 13 Gain Encoding Means 14 Distance calculation means 15 Linear prediction parameter decoding means 19 gain decoding means 20, 22 Pitch position extraction means 21, 23 Pitch synchronization means 24, 27 fixed codebook 25, 28 Second random codebook 26, 29 pitch position extraction means 30 Linear prediction parameter restoration means 31 Rise determination means 32 pitch extraction means 33 Adaptive code restoration means 34 Silent judgment means 35 Gain Restoration Means 36 Playback audio power control means 37 Voice state estimating means 38 noise generation means 39 Noise power control means

─────────────────────────────────────────────────── ─── Continuation of the front page (56) Reference JP 5-19794 (JP, A) JP 5-19796 (JP, A) JP 5-94200 (JP, A) JP 5- 108098 (JP, A) JP-A-6-202699 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) G10L 19/00-19/14 H03M 7/30 H04B 14/04

Claims (18)

(57) [Claims] 1. A voice encoding method for encoding input voice into a code of a plurality of parameters on a frame-by-frame basis, storing past drive excitation vectors and corresponding to adaptive codes.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector and a code vector that has a predetermined pitch synchronization position
Stores multiple codes and outputs a code vector corresponding to the noise code
Comprising a that noise codebook, any sign of the plurality of parameters except the noise code
Based on the
The pitch position is extracted from the sign position, and the pitch of each code vector is added to the extracted pitch position.
The pitch is set so that the sync position is matched.
Generates a second time-series vector that repeats Toll periodically
And, the first time series vector and the second time series vector
To generate the driving sound source vector, and drive the synthesis filter with the driving sound source vector
The play items, to evaluate the distortion for the input speech of said reproduced speech signals
A speech coding method characterized by determining a code. 2. A code of a plurality of parameters for each frame
Is a voice decoding method that decodes
Then, the past driving sound source vector is memorized, and it corresponds to the adaptive code.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector and a plurality of code vectors that have a predetermined pitch synchronization position are stored, and a noise codebook that outputs a code vector corresponding to a noise code is provided, and Which of the sign of the parameter
Based on the
The pitch position is extracted from the sign position, and the pitch of each code vector is added to the extracted pitch position.
The pitch is set so that the sync position is matched.
Generates a second time-series vector that repeats Toll periodically
And, the first time series vector and the second time series vector
The generated an addition to excitation vector, output sound by driving the synthesis filter with the excitation vector
A voice decoding method characterized by generating a voice . 3. A plurality of parameters for input voice in frame units
It is a voice encoding / decoding method that encodes the
Then, on the encoding side, the past driving excitation vector is stored, and it corresponds to the adaptive code.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector and a code vector that has a predetermined pitch synchronization position
Stores multiple codes and outputs a code vector corresponding to the noise code
Comprising a that noise codebook, any sign of the plurality of parameters except the noise code
Based on the
The pitch position is extracted from the sign position, and the pitch of each code vector is added to the extracted pitch position.
The pitch is set so that the sync position is matched.
Generates a second time-series vector that repeats Toll periodically
And, the first time series vector and the second time series vector
To generate the driving sound source vector, and drive the synthesis filter with the driving sound source vector
The play items, to evaluate the distortion for the input speech of said reproduced speech signals
The code is determined, and the decoding side stores the past driving excitation vector and corresponds to the adaptive code.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector and a code vector that has a predetermined pitch synchronization position
Stores multiple codes and outputs a code vector corresponding to the noise code
Comprising a that noise codebook, any sign of the plurality of parameters except the noise code
Based on the
The pitch position is extracted from the sign position, and the pitch of each code vector is added to the extracted pitch position.
The pitch is set so that the sync position is matched.
Generates a second time-series vector that repeats Toll periodically
Then, the first time series vector and the second time series vector are added to generate a driving sound source vector, and the synthesis filter is driven by the driving sound source vector to output the output sound.
A voice encoding / decoding method characterized by generating a voice. 4. A plurality of parameters are applied to the input voice for each frame.
It is a voice encoding method for encoding into a data code, storing past drive excitation vectors and corresponding to adaptive codes.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector, a predetermined pitch synchronization position, and a pitch period
The code vector with the cutout position determined accordingly.
Stores multiple codes and outputs a code vector corresponding to the noise code
Comprising a that noise codebook, from the feature position arranged at a pitch period interval in the current frame
The pitch position is extracted, and the above-mentioned code vectors are cut out according to the pitch period.
Position, and the pitch of the code vector
Periodic so that the period position matches the extracted pitch position above
To generate a driving sound source vector by adding the first time series vector and the second time series vector, and driving the synthesis filter with the driving sound source vector. Voice communication
The input signal of the reproduced audio signal above.
A speech coding method characterized in that distortion is evaluated to determine a code. 5. The audio encoding method according to claim 4, wherein the frame has an integral multiple of a pitch period. 6. The speech coding method according to claim 5, wherein a fixed length serving as a frame standard is determined, and an integer multiple of the pitch period is closest to the fixed length. 7. The speech coding method according to claim 5, wherein a fixed length serving as a frame standard is determined, and an average of integral multiples of the pitch period is always equal to or larger than the fixed length. 8. The voice encoding method according to claim 5, wherein a fixed point serving as a standard of a frame is determined, and an integer multiple of the pitch period is such that a frame boundary is closest to the fixed point. . 9. A code of a plurality of parameters for each frame
Is a voice decoding method that decodes
Then, the past driving sound source vector is memorized, and it corresponds to the adaptive code.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector, a predetermined pitch synchronization position, and a pitch period
The code vector with the cutout position determined accordingly.
Stores multiple codes and outputs a code vector corresponding to the noise code
Comprising a that noise codebook, from the feature position arranged at a pitch period interval in the current frame
The pitch position is extracted and the above code vectors are set to the pitch period.
Cut out at the corresponding cutout position, and this cut out code
The pitch sync position of the vector is the pitch position extracted above
Second time series vector that is periodically repeated to match
To generate the first time series vector and the second time series vector
To generate the driving sound source vector, and drive the synthesis filter with the driving sound source vector to output the output sound.
A voice decoding method characterized by generating a voice. 10. The speech decoding method according to claim 9, wherein the frame is an integral multiple of a pitch period. 11. A fixed length as a standard of the frame is defined,
The speech decoding method according to claim 10, wherein an integer multiple of the pitch period is closest to the fixed length. 12. A fixed length as a standard of the frame is defined,
11. The speech decoding method according to claim 10, wherein an average of integer multiples of the pitch period is always equal to or more than the fixed length. 13. A fixed point serving as a standard of a frame is defined,
11. The speech decoding method according to claim 10, wherein the integer multiple of the pitch period is such that a frame boundary is closest to the fixed point. 14. A plurality of parameters are applied to input voice for each frame.
It is a voice coding / decoding method that codes into the meter code.
Then, on the encoding side, the past driving excitation vector is stored, and it corresponds to the adaptive code.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector, a predetermined pitch synchronization position, and a pitch period
The code vector with the cutout position determined accordingly.
Stores multiple codes and outputs a code vector corresponding to the noise code
Comprising a that noise codebook, from the feature position arranged at a pitch period interval in the current frame
The pitch position is extracted, and the above-mentioned code vectors are cut out according to the pitch period.
Position, and the pitch of the code vector
Periodic so that the period position matches the extracted pitch position above
To generate a second time-series vector that is repeated, and to generate the first time-series vector and the second time-series vector.
To generate the driving sound source vector, and drive the synthesis filter with the driving sound source vector
The play items, to evaluate the distortion for the input speech of said reproduced speech signals
The code is determined, and the decoding side stores the past driving excitation vector and corresponds to the adaptive code.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector, a predetermined pitch synchronization position, and a pitch period
The code vector with the cutout position determined accordingly.
Stores multiple codes and outputs a code vector corresponding to the noise code
Comprising a that noise codebook, from the feature position arranged at a pitch period interval in the current frame
A second time when a pitch position is extracted, each of the code vectors is cut out at a cutout position corresponding to a pitch cycle, and the pitch synchronization position of the cut out code vector is periodically repeated so as to match the extracted pitch position. Generating a sequence vector, the first time series vector and the second time series vector
To generate the driving sound source vector, and drive the synthesis filter with the driving sound source vector to output the output sound.
A voice encoding / decoding method characterized by generating a voice. 15. A plurality of parameters are applied to input voice for each frame.
A speech coding apparatus that codes into a meter code, stores past drive excitation vectors, and supports adaptive codes.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector and a code vector that has a predetermined pitch synchronization position
Stores multiple codes and outputs a code vector corresponding to the noise code
Which is a random codebook or a code of the plurality of parameters excluding the random code.
Based on the
Pitch position extraction means for extracting the pitch position from the index position
And the pitch of each code vector above the extracted pitch position.
The pitch is set so that the sync position is matched.
Generates a second time-series vector that repeats Toll periodically
Pitch synchronization means, the first time series vector and the second time series vector
And an addition means for generating a driving sound source vector, and a synthetic filter for reproducing an audio signal with the driving sound source vector.
And the distortion of the reproduced voice signal with respect to the input voice
A distance calculating means for determining a sign.
Audio coding device. 16. A plurality of parameter marks for each frame
A voice decoding device that decodes a signal to generate an output voice.
Then, the past driving sound source vector is memorized, and it corresponds to the adaptive code.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector and a code vector that has a predetermined pitch synchronization position
Stores multiple codes and outputs a code vector corresponding to the noise code
Which is a random codebook or a code of the plurality of parameters excluding the random code.
Based on the
Pitch position extraction means for extracting the pitch position from the index position
And the pitch of each code vector above the extracted pitch position.
The pitch is set so that the sync position is matched.
Generates a second time-series vector that repeats Toll periodically
Pitch synchronization means, the first time series vector and the second time series vector
And adding means for generating a driving sound source vector, and a synthetic filter for generating an output sound by the driving sound source vector.
And a voice decoding device. 17. A plurality of parameters are applied to input voice for each frame.
A speech coding apparatus that codes into a meter code, stores past drive excitation vectors, and supports adaptive codes.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector, a predetermined pitch synchronization position, and a pitch period
The code vector with the cutout position determined accordingly.
Stores multiple codes and outputs a code vector corresponding to the noise code
From the random codebook and the characteristic positions that are lined up at pitch intervals in the current frame
Pitch position extracting means for extracting the pitch position , and a cutting position corresponding to the pitch period for each of the above code vectors
Position, and the pitch of the code vector
Periodic so that the period position matches the extracted pitch position above
To generate the second time series vector repeated
Averaging means, the first time series vector and the second time series vector
Means for generating a driving sound source vector, and a synthetic filter for reproducing an audio signal with the driving sound source vector
And the distortion of the reproduced voice signal with respect to the input voice
A distance calculating means for determining a sign.
Audio coding device. 18. A plurality of parameter marks for each frame
A voice decoding device that decodes a signal to generate an output voice.
Then, the past driving sound source vector is memorized, and it corresponds to the adaptive code.
The first time when the past driving sound source vector is periodically repeated
An adaptive codebook that outputs a sequence vector, a predetermined pitch synchronization position, and a pitch period
The code vector with the cutout position determined accordingly.
Stores multiple codes and outputs a code vector corresponding to the noise code
From the random codebook and the characteristic positions that are lined up at pitch intervals in the current frame
Pitch position extracting means for extracting the pitch position , and a cutting position corresponding to the pitch period for each of the above code vectors
Position, and the pitch of the code vector
Periodic so that the period position matches the extracted pitch position above
To generate the second time series vector repeated
Averaging means, adding means for adding the first time-series vector and the second time-series vector to generate a driving sound source vector, and a synthetic filter for generating an output sound by the driving sound source vector.
Speech decoding apparatus characterized by comprising a motor.